Operational Efficiency
How an on-premises chatbot transformed documentation access for a hospital group
Healthcare organizations need efficient access to internal knowledge, but documentation is often scattered across platforms, slowing teams and diverting focus from patient care. Agilytic partnered with a major hospital group to deploy an on-premises chatbot enabling teams to instantly retrieve information from internal documentation.
To protect confidentiality, we may alter specific details while preserving the accuracy of our core contribution.
Context and objectives
A leading hospital group managed a growing volume of internal documentation spread across multiple platforms, including Confluence (for internal wikis) and Jira (for ticketing). IT teams and staff members struggled to quickly locate the information they needed for daily operations.
Without an efficient knowledge retrieval system, the organization faced several operational challenges:
Time-consuming manual searches through extensive documentation repositories
Increased burden on subject matter experts who were repeatedly asked the same questions
Risk of inconsistent or outdated information being used in decision-making
Reduced productivity as employees spent time searching instead of focusing on core responsibilities
A critical requirement for this project was ensuring the solution complied with the healthcare sector's strict data privacy and security standards.
The project's objective was clear: build an on-premises chatbot that efficiently retrieves information from scattered internal documentation, while ensuring full compliance with data security requirements.
Approach
Designing a secure, on-premises architecture
Agilytic deployed a fully on-premises chatbot using an open-source platform with custom configurations tailored to the hospital group's secure infrastructure. The solution was built on a Retrieval-Augmented Generation (RAG) architecture, combining large language model capabilities with targeted information retrieval. A dual GPU setup ensured high-quality, multilingual responses.
Indexing and structuring internal knowledge
To enable fast and accurate retrieval, the team indexed multiple data sources into vector databases:
Several Confluence spaces containing internal documentation
Thousands of Jira tickets
Additional documents in PDF and DOCX formats stored within the team workspace
A specialized retrieval pipeline was created to handle content-based questions, addressed through RAG-powered semantic search.
Ensuring autonomy and long-term maintainability
Agilytic provided comprehensive documentation, technical handover sessions, and training materials to ensure the team could maintain and extend the solution independently.
Pipeline automation for knowledge base refresh scheduling was also implemented using Airflow.
Results
Agilytic deployed a fully on-premises chatbot using open-source technologies with strict data privacy compliance. The project delivered a solid proof of concept demonstrating the potential of AI-driven knowledge retrieval in a healthcare environment.
Key deliverables include:
On-premises chatbot with RAG architecture
Complete code repository with documentation
Architecture documentation including model configurations and files
Pipeline automation for knowledge base refresh scheduling
Technical handover sessions, training materials, user guide, and maintenance documentation
To safeguard confidentiality, we may modify certain details within our case studies.